Diagnosing Lung Cancer Earlier

Globally, lung cancer kills more than 1.5 million people a year, according to the World Health Organization. Compared with other high incidence cancers, it is a silent killer because very few symptoms are present in the early stages of the cancer process, which makes validating specific and sensitive screening challenging. Currently, late diagnoses, poor prognoses, and low survival rates are the norm.

In an effort to develop robust non-invasive early screening methods, the diagnostic potential of breath analysis has been investigated for a couple of decades; the hope is to isolate volatile organic compound (VOC) patterns that could be correlated to lung cancer. Despite our capability to isolate, separate, and identify VOCs, the lack of consistency between studies prevents breath analysis to be fully validated and used at the clinical level.

Our work in breath analysis is based on comprehensive two-dimensional gas chromatography (GC×GC) coupled to high-resolution (high-accuracy) time-of-flight mass spectrometry (TOF-MS), a technique that could be considered one of the most powerful tools in separation science. The high analytical resolution of GC×GC-HR-TOFMS relies on the proper combination of GC column phases for enhanced chromatographic resolution, but also on the efficiency of mass spectral deconvolution and the accuracy of the MS measurement. Such a multi-dimensional collection of data (1tR, 2tR, pure mass spectra, accurate mass) can improve our ability to differentiate between VOCs that find their origin inside the body versus VOCs issued externally by the patient and the environment.

In our approach, breath samples are collect using Tedlar (polyvinyl fluoride film) bags that we later empty into the thermal desorption tube. The sampling process is crucial, as the collection of human breath is subject to several interfering and confounding factors (food regime, smoking habits, and so on). Also, because exhaled air is a mixture of alveolar and ambient air, both endogenous substances and exogenous contaminants are sampled. Although lung washout using controlled medical air has been considered, our current strategy considers the collection of paired patient and control, together with environmental air samples and further comparison and subtraction between them. The idea is to minimize the effect of the presence of contaminants on class (patient versus control) segregation, while maintaining a simple and rapid sampling for patients.

After GC×GC-HR-TOFMS separation and tentative identification of breath VOCs, large data sets are organized in classes. Next, we apply several statistical tools and processes (Fisher ratios, principal component analyses, dispersion boxes) to reduce the data and extract a list of peaks that appear to be showing a certain level of specificity to samples included in the lung cancer class. We replicate measurements to create a composite chromatogram image that includes all compounds found in a given class. This data reduction allows the isolation of a limited set of peaks (n<20) from the original list of more than one hundred peaks. At this stage, the two-dimensional chromatograms are revisited to ensure that the highlighted peaks are properly shaped and that signals are free from chromatographic artifacts. The corresponding spectral information, including fragmentation patterns and accurate mass of parent ions, is then used to identify the peaks based on mass spectral library searching and molecular formulae calculation. These putative suspects can later be used to create a VOC profile that can be compared with VOC profiles of patients to contribute to diagnosing lung cancer sooner.

In parallel, we also developed an approach that is dedicated to trapping VOCs emitted by lung cancer cell cultures. Some preliminary tests have demonstrated that VOC profiles from the headspace of lung cancer cell media, isolated at various stages of cell growth, are different from VOC profiles of control cell cultures. Our aim is to see if these lung cancer cell culture head space VOCs can be compared, at least in terms of chemical families, to the putative biomarkers found in exhaled human breath.

In our view, the consideration of these two different aspects – breath analysis and cell culture headspace analysis – could enable better understanding of unknown pathophysiology that affects the VOC pattern in lung cancer. The use of state-of-the-art separation science tools for such an integrated approach will hopefully contribute to further demonstrating the early diagnostic potential of breath analysis – and get it one-step closer to clinical application.